Mistral Large Model

1 post · Feb 2024 – Mar 2026

What this trail covers

Mistral AI continues to evolve its flagship LLM with Mistral Large 2, featuring a 128k context window, enhanced multilingual capabilities, and improved reasoning and instruction-following. The model demonstrates competitive performance against leading models like GPT-4o and Claude 3 Opus, particularly in code generation and multilingual tasks. It also boasts improved function calling and retrieval skills. Mistral Large 2 is available via la Plateforme, cloud service providers (GCP, Azure, AWS, IBM), and its weights are released under the Mistral Research License for non-commercial use, with commercial licenses available.

All Mistral AI trails

Mar 2026Introducing Mistral Small 4 | Mistral AIIntroduces Mistral Small 4, a new hybrid model that unifies instruct, reasoning, and multimodal capabilities. Key architectural details include a Mixture of Experts (MoE) with 128 experts (4 active per token), 119B total parameters (6B active), and a 256k context window. It features configurable reasoning effort ('none' for low-latency, 'high' for deep reasoning) and native multimodality (text and image inputs). Performance highlights include a 40% reduction in end-to-end completion time and a 3x increase in requests per second compared to Mistral Small 3. The model is open-source under Apache 2.0 and optimized for NVIDIA hardware, with availability on various platforms including Mistral API, AI Studio, and NVIDIA NIM.
Jul 2024Large Enough | Mistral AIThis post introduces Mistral Large 2, detailing its 128k context window, extensive language support (including 80+ coding languages), and 123 billion parameters designed for single-node inference. It highlights performance gains on MMLU (84.0% accuracy), code generation benchmarks (outperforming previous Mistral Large and on par with GPT-4o, Claude 3 Opus, Llama 3 405B), and reasoning benchmarks (GSM8K, MATH) with a focus on reducing hallucinations and acknowledging knowledge gaps. Instruction following and conversational capabilities are improved, with a focus on succinct responses for business applications. Multilingual performance on MMLU is showcased, and enhanced function calling and retrieval skills for complex business applications are noted. Mistral Large 2 is available on la Plateforme as `mistral-large-2407`, with weights released under the Mistral Research License for research and non-commercial use. Fine-tuning capabilities are extended to Mistral Large, Nemo, and Codestral on la Plateforme. Partnerships with Google Cloud Platform (Vertex AI), Azure AI Studio, Amazon Bedrock, and IBM watsonx.ai are expanded.
Feb 2024Au Large | Mistral AIThis post announces Mistral Large, detailing its multilingual fluency, 32K context window, precise instruction following, and native function calling capabilities. It also introduces Mistral Small for low-latency workloads and highlights their availability on La Plateforme and Azure. Function calling and JSON format are specifically mentioned as features for Mistral Small and Large.
Advertisement